Posts
Comments
I just cleaned up all the formatting and converted the LaTeX. Sorry for the inconvenience!
I was already wondering about that. Will move this post back to drafts.
Both alt-text and captions can be set in the new editor! (Can be activated by opting into beta features on your profile settings)
Looks like I missed one. How about now?
Oops, I just created another copy of this. Will draft this one for now.
Nope, looks like a bug, will look into it.
Should be fixed now, I think
(Mod note: I added spoiler tags to both this comment and the OP to protect the eyes of the innocent frontpage scrollers who might have not read HPMOR)
I really like the FB crossposts here, and also really like this specific comment. Might be worth polishing it into a top-level post, either here or on the EA Forum sometime.
TypeKit, aka Adobe Fonts, is a font library and font hosting service. We could probably consolidate this with one of the other CDNs being used, but font-hosting involves some user-agent-string based compatibility polyfills, which would be somewhat annoying to reproduce ourselves.
Small correction to this. As I mentioned below, we don't actually have a license to host the fonts we are serving ourselves. We could buy one, but it would probably run into at least hundred and possibly thousands of dollars per year, because fonts are expensive.
Jsdelivr provide marginal benefit at some cost, which aren't worth it in my opinion.
The only place where we use jsdelivr is for serving MathJax, for which it is the canonical source that Mathjax links to in the documentation, which seems good because it allows people to cache Mathjax for multiple sites, so I think this is the best solution here. Seems worse for us to set up our own CDN, and worse for it to be served from the LessWrong server, since that makes our job harder.
Typekit.net (presumably) provides the fonts used, which is useful for caching although LW could also do it locally.
Small comment on this: We have a Typekit subscription, which sadly does not actually allow us to download the fonts and serve them locally. We have to serve them directly from Adobe's servers. It's slightly annoying, but I don't think it's bad enough that I would want to stop using Typekit (which has overall been pretty decent and gives us access to a really wide range of fonts).
Yep, this is basically right.
We have recently experimented with LogRocket, but it's currently deactivated (though we might activate it again in the future). People should also feel free to block it, since the benefit for us is just from getting more data on how on-average users interact with the site.
We don't use dropbox, and indeed it isn't loaded on any page I could find. People often use Dropbox to host images for LessWrong posts, so my guess is that where that request came from. The same goes for dl.drop and the dropboxusercontent URL.
Google Analytics is just really useful. We are building up internal analytics infrastructure, but I think we are still quite a bit away from being able to shut down Google Analytics.
The Googleapis are likely for Google's ReCaptcha which we use to identify bots.
Overall, feel free to deactivate basically all of them, and nothing horrible should break. With the exception of typekit, algolia, jsdelivr (which would break LaTeX editing) and cloudflare. You can just deactivate Intercom in your user settings if you don't want it, but you can also just block requests to Intercom.com.
This is the first time that I feel like I've actually seen someone use the four knowledge terms ("episteme", "techne", "metic" and "gnosis") in a way that felt like it created a coherent picture. I got used to thinking about "metis" as distinct from other types of knowledge, but this has also helped me place where "techne", "episteme" and "gnosis" fall.
I have lots of thoughts, but most of all I am just really grateful for you writing this retrospective. I think it's a really great public service and I've already been thinking a lot about this post since I read it the first time a few hours ago, and expect to continue thinking about it a lot more.
Yeah, the internet archive is doing some stuff that is a bit hard to work around with our architecture (it tries to execute our javascript bundle and then sends requests to nonexistent servers). It's been on my to-do list to fix for a while (probably by just somehow telling the archive to just serve the plain HTML versions of the pages that we send, or just making sure that our javascript bundle fails more gracefully when it gets called in the ways the internet archive is doing it).
Until then, backing up the greaterwrong mirror seems like something that gets us most of the value, though I do really want to get around to fixing the archive.org versions for the main site.
I think various discussions on LessWrong are probably your best bet. A lot of LessWrong discussion is distributed over a large number of posts and platforms, so things end up less centralized than for SSC stuff, which has benefits and drawbacks.
For an experience similar to an SSC Open Thread, I think the LW Open Threads are your best bet, though they are definitely a lot less active than the SSC ones.
"Synchronized moderating" could be an olympic sport I guess :P (We both wrote a reply with functionally the same content at the same time)
Something in this space seems pretty plausible to me. We are always monitoring contributions from new users, so I think we would notice relatively quickly, and I agree that as soon as we see a suspicious uptick we might want to limit contributions, but I do think I would want to wait until we see the initial signs.
Welcome limestone! And feel free to leave comments here or ping the admins on Intercom (the small chat bubble in the bottom right) if you run into any problems!
Yeah, that seems reasonable. I will move this back to drafts.
Yep, all auth currently happens via Meteor.
We sadly don’t really have any infrastructure set up to hand out programmatic auth tokens, but I think we set the expiration date to something like 5 years, so I don’t think you should run into much of an issue.
And yeah, Meteor generally communicates over websockets. So my guess is that includes the auth part.
I would recommend just inspecting a graphql query you send while you are logged in and make a request, using your browsers network tools. Chrome has a great utility of copying any query you sent in a bunch of script-ready formats, like a call to cURL or fetch, including any headers you sent along.
Glad to hear that! Seems like a good reason to publish this as a top-level post. Might go ahead and do that in the next few days.
They will get a notification that someone shared them on a draft, and the post will show up below their own drafts on their personal profile page (only visible to them, obviously). However, until we ship some of the inline-commenting features we've been working on, there isn't actually a way for them to leave feedback on the post, which is unfortunate (we used to allow users to leave comments on drafts, but there were various problems with that, among them it being unclear whether those comments should persist to the published version of the post or not).
So that means that while they will get notifications about being shared, you should probably PM them if you want them to leave feedback. We do have an editor in beta that will allow users who are shared on a draft to leave Google Docs style inline comments, which I expect to go up within the next few months.
(We inherited the terms of use from the old LessWrong so while I tried my best to understand them, I don't have as deep of a sense of understanding as I wish I had, and it seemed more important to keep the terms of use consistent than to rewrite everything, to maintain consistency between the agreements that authors made when they contributed to the site at different points in time)
The Wikipedia article on Moral Rights:
The moral rights include the right of attribution, the right to have a work published anonymously or pseudonymously, and the right to the integrity of the work.[1] The preserving of the integrity of the work allows the author to object to alteration, distortion, or mutilation of the work that is "prejudicial to the author's honor or reputation".[2] Anything else that may detract from the artist's relationship with the work even after it leaves the artist's possession or ownership may bring these moral rights into play. Moral rights are distinct from any economic rights tied to copyrights. Even if an artist has assigned his or her copyright rights to a work to a third party, he or she still maintains the moral rights to the work.[3]
What exactly these rights are seems to differ a bunch from country to country. In the U.S. the protection of these moral rights is pretty limited. From the same article:
Moral rights[17] have had a less robust tradition in the United States. Copyright law in the United States emphasizes protection of financial reward over protection of creative attribution.[5]:xiii The exclusive rights tradition in the United States is inconsistent with the notion of moral rights as it was constituted in the Civil Code tradition stemming from post-Revolutionary France. When the United States acceded to the Berne Convention, it stipulated that the Convention's "moral rights" provisions were addressed sufficiently by other statutes, such as laws covering slander and libel.[5]
Concrete instances where I can imagine this waiving becoming relevant, and where I think this makes sense (though this is just me guessing, I have not discussed this in detail with a lawyer):
- An author leaves a comment on a post that starts with a steelman of an opposing position. We display a truncated version of the comment by default, which now only shows them arguing for a position they find abhorrent. This could potentially violate their moral rights by altering their contribution in a way that violates their honor or reputation.
- An author leaves a comment and another user quotes a subsection of that comment, bolding, or italicizing various sections that they disagree with, and inserting sentences using notation.
All remaining problems with document submission should be resolved. If you had opted into beta features and had trouble submitting documents in the past few hours, you should be able to do that again, and please let me know via Intercom if you can't.
We might still have some problems with comment and PM submissions that I am looking into. Not sure what's causing that.
Sorry for the outages today (we had two outages, one around 1:00PM PT, one around 3:30PM PT, with intermittent slow requests in the intervening time). As far as I can tell it was caused by a bot that was crawling particularly expensive pages (pages with tons of comments) at a relatively high rate. We've banned the relevant IP range and everything appears back to normal, though I am still watching the logs and server metrics attentively.
Again, sorry for any inconveniences this caused, and please let us know via Intercom if you run into any further problems.
Can confirm. We have them all over our house and they've significantly reduced wire-associated mess.
Alas, I tried to fix it for the author, but it's not super easy. We don't (yet) have first-class support for footnotes in the WYSIWYG editor, so I can't really make the footnotes work in the way they were probably intended.
Just to clarify, I wouldn't really say that "we are building our own editor". We are just customizing the CKEditor 5 framework. It is definitely a bunch of work, but we aren't touching any low-level abstractions (and we've spent overall more time than that trying to fix bugs and inconsistencies in the current editor framework we are using, so hopefully it will save us time in the long-run).
Promoted to curated: This feels like the best summary of all the key Covid updates to make over the last month or so, and generally just a post with a lot of good points.
I generally don't want all of LessWrong to become about Covid, but creating common knowledge of things like this feels like the right tradeoff in handling the visibility of Covid stuff, though I am not super confident.
Mod note: I decided to move the second half of this comment to the Open Thread, because Zvi explicitly requested that comments should stay on-topic. Here is a link to the new comment.
I think it's a decent thing to do, and I've done the same myself a few years ago before I attended a workshop. I think it was reasonably useful. I did also try to teach a lot of the material in the handbook and on LessWrong to my friends, which I found a lot more useful for actually understanding the stuff.
Huh, do you have a link for this claim? In particular the sentence "the lowest % and worst performance of all groups asked to predict the result"?
Alas, it isn't out on Mac, so I can't play it, but I will give DreamQuest a shot, which I haven't heard about until now.
Welcome! I hope you will enjoy the things you find here!
Mini-feature announcement I didn't have anywhere else to put: Links to Metaculus forecasting questions now have hover-previews that allow you to see the state of the prediction on metaculus. See this link for an example.
Note: This post was originally posted to the DeepMind blog, so presumably the target audience is a broader audience of Machine Learning researchers and people in that broad orbit. I pinged Vika about crossposting it because it also seemed like a good reference post that I expected would get linked to a bunch more frequently if it was available on LessWrong and the AIAF.
Note: We don't have formal rules about this since this has never come up, but I think questions should actually have something concrete about the question in the title, and should avoid being names something like "True or False" or "Is it true?" or anything in that reference class.
You can remove tags by downvoting their relevance below 1.
Promoted to curated. My engagement with this post was interesting, I went from "this post makes a trivial point" to "this post is obviously wrong" to "I am really confused by this post" to "this post is obviously right and makes some counterintuitive claims". I am not really sure what I initially thought this post was saying, so maybe I just confused myself, but I do think this journey of mine is pretty significant evidence that the post said something interesting.
I also really like having a reference that covers how to deal with 50% predictions pretty comprehensively, which this post does.
My guess is that it was an accident. Fixed it for the author.
(Note: I already created the tag)
I am worried about "Frameworks" being too broad. Like, isn't every model a framework of some kind?
Seems like a great idea, I created it here: https://www.lesswrong.com/tag/financial-investing
Promoted to curated: This is a technique I've seen mentioned in a bunch of places, but I haven't seen a good writeup for it, and I found it quite valuable to read.
Yeah, I really don't know yet how we should handle hierarchical tagging. My current model is that "That Alien Message" should just be tagged both AI Alignment and AI Boxing, and that we should be hesitant to have too much formal hierarchy in the tags, since I expect that most relationships will turn out to be only "almost hierarchical" in that it's not literally always the case the we would want to apply both tags together, and being forced to always have them come in a package is I think pretty costly.
I do think the description of the AI Boxing tag should link to the AI Alignment tag, so that should be edited in. And in general related tags should link to each other a lot, to make it easier to discover content that way.
"Rationality" is part of a special class of 5 tags that are designed to be general, and be primarily used for broad filtering purposes (i.e. the 5 core tags). Agree that it would be far too broad for most tags.
Re. "Problem-Solving Tactics": I agree that something in that space is good. We currently have the "Techniques" tag, which is pretty similar to that, so I added it to that tag. Do you think we want something again more specific than that?